Blinding, Blondlot and N-rays

Dr René-Prosper Blondlot was a French physicist who announced in 1903 that he had discovered a new form of radiation. X-rays had been discovered to great acclaim in 1895 by Wilhelm Röntgen, who went on to win the Nobel Prize for his discovery. Blondlot was inspired by Röntgen’s discovery and he determined to make as great a name for French physics as Röntgen had for Germany. With a nod to Röntgen and history, Blondlot named his discovery “N-rays.” His findings were replicated by many French scientists but could not be reproduced by physicists in England, Germany or the US.

Blondlot’s N-ray photographs [from rexresearch.com]

Robert W. Wood, a skeptical American physicist whose own attempts at replicating the N-ray experiment had revealed nothing, arranged to visit Blondlot’s laboratory for a demonstration. Blondlot and his assistant ran the experiment for Wood’s benefit. The apparatus was a complicated affair involving a hot wire inside an iron tube that generated the N-rays. The N-rays were then deflected by an aluminium prism so that they shone onto a thread of calcium sulphide and gave the thread a subtle glow.

The account of what followed has many different variations, partly because Wood himself offered different versions of events over many years, but this summarises what everyone agrees on: Unbeknownst to his hosts, Wood secretly removed the aluminium prism from the apparatus. Blondlot still perceived the glow on the thread. Then Wood replaced the original aluminium prism. Blondlot’s assistant caught him at it and informed Blondlot, who assumed that the prism was missing and observed that the glow had disappeared from the thread…but Wood had replaced the essential component by then. Blondlot was seeing what he expected to see, not what was there.

The story of Blondlot and N-rays is one of the cautionary tales of science — or to use Christine Barry’s terminology, the narrative fable of Blondlot’s N-rays has become a standard rhetoric in the scientific hegemony and is used politically to exclude heterodox epistemologies. But it is more than just one story. There are dozens of other examples, from Pons and Fleischman’s cold fusion experiment or facilitated communication, and any one of them could stand in as the standard fable. The N-ray story persists because illustrates a fundamental principle. Researchers delude themselves.

The lesson of self-delusion had a massive effect on clinical trial designs. Scientists learned that if you let the patient know which treatment they were getting, it could skew the results, but they also learned that the researchers themselves were just as prone to skewing results. This was not fraudulent behaviour. It was completely unconscious. Some of biases tended to make a therapeutic effect more prominent (subjective reporting tends to favour therapies that are exciting and new). Other biases can hide a real effect (doctors tend to allocate sicker people to exciting new treatments because, being sicker, they are in greater need). The ideal way to test a therapy is to “blind” the participants so that they don’t know what they are getting. When the researchers themselves are not aware of the treatment given until the allocation is revealed at the end of the trial, this is called “double blinding.” It is now the standard design for clinical trials, and the reason why it has become the gold standard is that scientists have at long last recognised their own capacity to fool themselves.

This lesson has not carried across to all fields of academia. In personal communication, I have often heard veterinarians say they have no need for blinded trials because animals don’t experience the placebo effect. This is, of course, utter rubbish. Veterinarian work involves animals that exist in a social relationship with other animals, including their human handlers. Whether the animal is a pet, a seeing-eye dog, or a dairy cow, it will respond to its handlers and to the other animals around it. If the handlers treat them differently, they will respond differently.

In 1935, Ella Hediger and Harry Gold published a paper examining the effects of two types of ether as surgical anaesthetic. Hediger and Gold went to great lengths to hide the type of ether from the anaesthetists. They put them in identical bottles. They rotated them on a daily basis. They randomised the rotation to prevent anaesthetists figuring out which ether they were using. The reason they went to all this trouble was:

We realize that the subjective elements involved in the classification of anesthesias as “satisfactory” or “unstatisfactory” introduce variable factors with a considerable margin of error and that under ordinary circumstances opinions which are based on such estimates are vitiated by the bias of the anesthetist who knows the source of the ether. The significant point of this study is that the general appraisals were made entirely free of possible prejudices and preconceived notions regarding the relative value of different types of ether, since those who made them had no knowledge of the source of the ether.

If a study on unconscious patients needs to be double-blinded, and if the “hard” physical evidence for N-rays turned out to be a perceptual phantom, then veterinarians who are blithely confident of their powers of observation ought to be giving more thought to the matter. Fortunately, veterinary scientists are fully of aware of the value of blinding and the research literature is full of well-designed, prospective double-blind RCTs. Naturally, veterinarian scientists are as interested in the possibilities of homeopathy as medical scientists. Here is F. J. van Sluijs, a Dutch veterinarian scientist, from a 2004 article summarising the evidence from a large number of trials and meta-analyses:

The golden standard is the prospective randomized double-blind clinical trial. For medicine this is a formidable enough challenge; for homeopathy it appears to be insurmountable. After seven years of investigation costing more than $100 million per year, there is still no evidence whatever that extremely diluted solutions of homeopathic substances have any effect.

Perhaps the most interesting observation that van Sluijs makes is that the benefits of homeopathy become harder and harder to observe as the quality of the trial design improves, to the point where benefits disappear altogether in “gold standard” trials.

Of proofs and puddings

When I was a young hospital resident, I witnessed firsthand the capacity for randomised control trials to overturn accepted knowledge. Within a few years of each other, two trials were published that revolutionised the treatment of heart attacks. The essential pathology of heart attacks is very simple: a clot forms in a damaged heart artery; the clot breaks off and is carried by blood-flow down the artery, which narrows as it branches, until the clot comes to a place where the artery is too narrow to pass. The clot becomes lodged in the artery and blocks the bloodflow to the more distant arms of the artery. Heart muscle needs a constant wash of oxygen to keep beating, so a sudden blockage of the blood supply leads almost immediately to pain, soon to loss of function, and within minutes to hours (depending on the location and severity of the blockage) irreversible cell death.

In the late 1980s and early 1990s, heart disease was treated a great deal differently to how it is managed today. Two RCTs played a big part in that turnaround: the CAST trial and the ISIS-2 trial.

One of the most promising strategies for treating heart attacks was dissolving the clot before permanent damage occurs. It certainly made sense. The problem was that while there were effective treatments to prevent clotting in the first place (warfarin, heparin, and aspirin), there was only one available treatment to dissolve an existing clot. This treatment is called streptokinase (now there are more treatments available). Its name is a clue to its origin: it was first isolated in streptococcus bacteria. To streptococci, this protein was part of its invasive armoury. Clotting is part of the body’s defence not just against bleeding but also against invading bacteria, and streptokinase allowed the bacteria to chew through solid walls of the fibrin that holds together blood clots. To use a mediaevel analogy, fibrin is part of the body’s castle wall against invaders, and streptokinase is a bacterial siege engine.

While the use of streptokinase looked promising in theory, there were two big arguments against its use. The first was that dissolving clots is not necessarily beneficial. Most patients with heart attacks also have risk factors for strokes, and since about a third of strokes are caused by uncontrolled bleeding into the brain, it is risky to go about dismantling the natural defence against such haemorrhagic strokes. The second fear was that streptokinase, being a bacterial protein, would cause massive allergic responses that could also be fatal. The first trial of intravenous streptokinase for heart attacks took place way back in 1958, but these concerns stopped it from becoming standard treatment.

In the ISIS-2 trial, patients attending emergency departments for acute heart attacks were randomised to receive streptokinase, aspirin, both, or neither. The results showed that using streptokinase and aspirin together was most effective: the odds of dying immediately after a heart attack dropped by a massive 42% while the risks of haemorrhagic stroke and allergic hypersensitivity were not increased. Conventional wisdom was wrong. (The original ISIS-2 paper is available online in abstract only, but the ten-year follow-up study published in 1998 is available in full text.)

15-month survival curve from the ISIS-2 trial.

10-year survival curve from ISIS-2.
[This graph from Baigent et al, BMJ 316:1337-1343, 1998; previous graph adapted from same.]

The CAST trial was even more dramatic. In the first few days after a heart attack, the most common cause of death is arrhythmia, which means the oxygen-starved heart can go into abnormal rhythms. Instead of beating regularly, the heart can go into paroxysms which reduce its ability to perform as an efficient pump. With some rhythms such as ventricular fibrillation, the heart’s efficiency drops to zero. It flops around like a fish on a boat instead of pumping blood, and when the heart stops pumping blood, the patient dies. So it made sense to prevent arrhythmias. The heart itself may be just as damaged in the long run, but at least the patient would survive the first dangerous days after a heart attack. A well-received 1998 study known as the CAPS trial showed that anti-arrhythmics such as flecainide were very effective at stopping arrhythmias, but nobody had tested their capacity to save lives until the CAST trial (abstract available), published in a series of papers over 1991-3.

The CAST trial was never completed. Midway through the study, it became obvious that the drugs may have prevented arrhythmias but they were killing patients. In fact, patients were twice as likely to die within the first year after a heart attack if they received anti-arrhythmic medication.

Survival curve from the CAST trial. [Adapted from Echt et al, NEJM 324:781-788, 1991.]

It is studies like these that make it impossible for Christine Barry to completely “deconstruct” the use of RCTs in medicine. But Barry is not trying to denigrate the use of RCTs in orthodox medicine; she is trying to protect alternative therapists from their corrosive influence.

(Next: Demolishing the deconstruction…)